Red, Red Wine
Jason Horowitz
June 12, 2018
## ── Attaching packages ───────────────────────────────────────────────────── tidyverse 1.2.1 ──## ✔ ggplot2 2.2.1 ✔ purrr 0.2.5
## ✔ tibble 1.4.2 ✔ dplyr 0.7.5
## ✔ tidyr 0.8.1 ✔ stringr 1.3.1
## ✔ readr 1.1.1 ✔ forcats 0.3.0## ── Conflicts ──────────────────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ dplyr::arrange() masks plyr::arrange()
## ✖ dplyr::combine() masks gridExtra::combine()
## ✖ purrr::compact() masks plyr::compact()
## ✖ dplyr::count() masks plyr::count()
## ✖ dplyr::failwith() masks plyr::failwith()
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::id() masks plyr::id()
## ✖ dplyr::lag() masks stats::lag()
## ✖ dplyr::mutate() masks plyr::mutate()
## ✖ dplyr::rename() masks plyr::rename()
## ✖ dplyr::summarise() masks plyr::summarise()
## ✖ dplyr::summarize() masks plyr::summarize()##
## Attaching package: 'GGally'## The following object is masked from 'package:dplyr':
##
## nasa## [1] "/Users/jasonhorowitz/Desktop/DAND/t2-p2-project-EDA"Intro
The objective of this project is to perform EDA (exploratory data anlalysis) examining relationships among variables in the red wine dataset. This is one of two datasets presented by Udacity which relate to red and white variants of Portuguese Vinho Verde wine.
There are a total of 1599 observations (wines) in the dataset displaying twelve variables including eleven input variables (based on physicochemical tests) and one output variable based on sensory data, a quality score.
The physico-chemical variables are (descriptive text from text file included with dataset):
- fixed acidity (tartaric acid - g / dm^3): most acids involved with wine are fixed or nonvolatile (do not evaporate readily).
- volatile acidity (acetic acid - g / dm^3): the amount of acetic acid in wine, which at too high of levels can lead to an unpleasant, vinegar taste.
- citric acid (g / dm^3): found in small quantities, citric acid can add ‘freshness’ and flavor to wines.
- residual sugar (g / dm^3): the amount of sugar remaining after fermentation stops, it’s rare to find wines with less than 1 gram/liter and wines with greater than 45 grams/liter are considered sweet.
- chlorides (sodium chloride - g / dm^3): the amount of salt in the wine
- free sulfur dioxide (mg / dm^3): the free form of SO2 exists in equilibrium between molecular SO2 (as a dissolved gas) and bisulfite ion; it prevents microbial growth and the oxidation of wine.
- total sulfur dioxide (mg / dm^3): amount of free and bound forms of S02; in low concentrations, SO2 is mostly undetectable in wine, but at free SO2 concentrations over 50 ppm, SO2 becomes evident in the nose and taste of wine.
- density (g / cm^3): the density of wine is close to that of water depending on the percent alcohol and sugar content.
- pH: describes how acidic or basic a wine is on a scale from 0 (very acidic) to 14 (very basic); most wines are between 3-4 on the pH scale
- sulphates (potassium sulphate - g / dm3): a wine additive which can contribute to sulfur dioxide gas (S02) levels, wich acts as an antimicrobial and antioxidant.
- alcohol (% by volume): the percent alcohol content of the wine.
- quality (score between 0 and 10):the median of at least 3 evaluations made by wine experts with grading between 0 (very bad) and 10 (very excellent).
Univariate Plots Section
Let’s take a look at the structure of the dataset and let’s also summarize the variables:
## 'data.frame': 1599 obs. of 13 variables:
## $ X : int 1 2 3 4 5 6 7 8 9 10 ...
## $ fixed.acidity : num 7.4 7.8 7.8 11.2 7.4 7.4 7.9 7.3 7.8 7.5 ...
## $ volatile.acidity : num 0.7 0.88 0.76 0.28 0.7 0.66 0.6 0.65 0.58 0.5 ...
## $ citric.acid : num 0 0 0.04 0.56 0 0 0.06 0 0.02 0.36 ...
## $ residual.sugar : num 1.9 2.6 2.3 1.9 1.9 1.8 1.6 1.2 2 6.1 ...
## $ chlorides : num 0.076 0.098 0.092 0.075 0.076 0.075 0.069 0.065 0.073 0.071 ...
## $ free.sulfur.dioxide : num 11 25 15 17 11 13 15 15 9 17 ...
## $ total.sulfur.dioxide: num 34 67 54 60 34 40 59 21 18 102 ...
## $ density : num 0.998 0.997 0.997 0.998 0.998 ...
## $ pH : num 3.51 3.2 3.26 3.16 3.51 3.51 3.3 3.39 3.36 3.35 ...
## $ sulphates : num 0.56 0.68 0.65 0.58 0.56 0.56 0.46 0.47 0.57 0.8 ...
## $ alcohol : num 9.4 9.8 9.8 9.8 9.4 9.4 9.4 10 9.5 10.5 ...
## $ quality : int 5 5 5 6 5 5 5 7 7 5 ...## X fixed.acidity volatile.acidity citric.acid
## Min. : 1.0 Min. : 4.60 Min. :0.1200 Min. :0.000
## 1st Qu.: 400.5 1st Qu.: 7.10 1st Qu.:0.3900 1st Qu.:0.090
## Median : 800.0 Median : 7.90 Median :0.5200 Median :0.260
## Mean : 800.0 Mean : 8.32 Mean :0.5278 Mean :0.271
## 3rd Qu.:1199.5 3rd Qu.: 9.20 3rd Qu.:0.6400 3rd Qu.:0.420
## Max. :1599.0 Max. :15.90 Max. :1.5800 Max. :1.000
## residual.sugar chlorides free.sulfur.dioxide
## Min. : 0.900 Min. :0.01200 Min. : 1.00
## 1st Qu.: 1.900 1st Qu.:0.07000 1st Qu.: 7.00
## Median : 2.200 Median :0.07900 Median :14.00
## Mean : 2.539 Mean :0.08747 Mean :15.87
## 3rd Qu.: 2.600 3rd Qu.:0.09000 3rd Qu.:21.00
## Max. :15.500 Max. :0.61100 Max. :72.00
## total.sulfur.dioxide density pH sulphates
## Min. : 6.00 Min. :0.9901 Min. :2.740 Min. :0.3300
## 1st Qu.: 22.00 1st Qu.:0.9956 1st Qu.:3.210 1st Qu.:0.5500
## Median : 38.00 Median :0.9968 Median :3.310 Median :0.6200
## Mean : 46.47 Mean :0.9967 Mean :3.311 Mean :0.6581
## 3rd Qu.: 62.00 3rd Qu.:0.9978 3rd Qu.:3.400 3rd Qu.:0.7300
## Max. :289.00 Max. :1.0037 Max. :4.010 Max. :2.0000
## alcohol quality
## Min. : 8.40 Min. :3.000
## 1st Qu.: 9.50 1st Qu.:5.000
## Median :10.20 Median :6.000
## Mean :10.42 Mean :5.636
## 3rd Qu.:11.10 3rd Qu.:6.000
## Max. :14.90 Max. :8.000Copying Dataframe and Transforming/Adding Variables
The first thing we are going to do is to copy the rw dataframe so that subsequent transformations do not affect the original df. Since it will be useful to have the quality variable in both factor and integer from, we will create a factorized version of the quality variable (quality_factor). Also we will transform the X variable into a factor (since statistically summarizing this variable is meaningless) and will rename it to the more humanly readable “id”.
## 'data.frame': 1599 obs. of 14 variables:
## $ id : Factor w/ 1599 levels "1","2","3","4",..: 1 2 3 4 5 6 7 8 9 10 ...
## $ fixed.acidity : num 7.4 7.8 7.8 11.2 7.4 7.4 7.9 7.3 7.8 7.5 ...
## $ volatile.acidity : num 0.7 0.88 0.76 0.28 0.7 0.66 0.6 0.65 0.58 0.5 ...
## $ citric.acid : num 0 0 0.04 0.56 0 0 0.06 0 0.02 0.36 ...
## $ residual.sugar : num 1.9 2.6 2.3 1.9 1.9 1.8 1.6 1.2 2 6.1 ...
## $ chlorides : num 0.076 0.098 0.092 0.075 0.076 0.075 0.069 0.065 0.073 0.071 ...
## $ free.sulfur.dioxide : num 11 25 15 17 11 13 15 15 9 17 ...
## $ total.sulfur.dioxide: num 34 67 54 60 34 40 59 21 18 102 ...
## $ density : num 0.998 0.997 0.997 0.998 0.998 ...
## $ pH : num 3.51 3.2 3.26 3.16 3.51 3.51 3.3 3.39 3.36 3.35 ...
## $ sulphates : num 0.56 0.68 0.65 0.58 0.56 0.56 0.46 0.47 0.57 0.8 ...
## $ alcohol : num 9.4 9.8 9.8 9.8 9.4 9.4 9.4 10 9.5 10.5 ...
## $ quality : int 5 5 5 6 5 5 5 7 7 5 ...
## $ quality_factor : Factor w/ 6 levels "3","4","5","6",..: 3 3 3 4 3 3 3 5 5 3 ...Structure and Summary of Data
The original dataset consists of 12 variables (plus an index (X) which identifies the different wines). The variables are all number or integer data types. There are a total of nearly 1600 observations in the dataset. There are no missing values and each row of the database equals one observation: so the dataset is both clean and tidy.
Histogram of Wine Quality Scores
For our first histogram, we will plot the wine quality scores. It is notable that greater than 75% of the quality scores are concentrated in the middle (5-6) of the score range (1-10). The shape of the curve is normal-ish.
Also worth noting is that the Quality score is an integer data type despite the description that it is derived from observations of at least 3 experts. It would have been nice to have the precision of a decimal average for quality scores. On the other hand, it simplifies things by making a bin = 1 optimal for displaying the distribution.
In fact, since the ratings are distributed across six integers (3-8), it made sense to convert this into a factor variable. The results coincide with the following qualitative (subjective) ratings:
- 3 = “Awful”
- 4 = “Poor”
- 5 = “Fair”
- 6 = “Good”
- 7 = “Very Good”
- 8 = “Superior”
Following is the summary count of wines by quality score.
## 3 4 5 6 7 8
## 10 53 681 638 199 18Histograms of Other Variables
Let’s look at histograms of the other variables adjusting the binwidth as appropriate to have a look at the distribution of values. I will also perform transformations on plots that require normalization.
I hope that plotting variables and some minor additional research into them will give us ideas for further investigation in the bivariate and multivariate analysis to follow.
Fixed Acidity
Since this appears to have a long-tail distribution, we will transform the variable using the log base 10 scale.
With this transformation, the distribution of the data follows a more normal distribution.
Volatile Acidity
The distribution for volatile acidity also seems to be longtailed - and with some outliers. Let’s perform a similar log10 transformation - also eliminating the top and bottom 1% of the values.
## Scale for 'x' is already present. Adding another scale for 'x', which
## will replace the existing scale.
As with fixed acidity, the transformed volatile acidity data follows a somewhat more normal distribution.
Citric Acid
A signicant number of wines have citric acid near 0. Let’s check how many actually have no trace of citric acid.
## [1] 132 14## [1] 0.08255159So 132 items (8.2% of the observations) contain no trace of citric acid. This makes sense because the citric acid most usally found in wine is a supplement added after primary alcohol fermentation (citric acid is present only in minute quantities in grapes) In the European Union, use of citric acid for acidification is prohibited, but limited use of citric acid is permitted for removing excess iron and copper from the wine if potassium ferrocyanide is not available. Wikipedia
Also note - Citric acid imparts a citric character that enhances the taste of many white and blush wines. However, citric acid is seldom used in red wine. The distinctive citric taste may not be appropriate for many types of red wine. In addition, the risk of biological instability is much greater in red wines.Lum Eisenman, The Home Winemaker’s Manual
The description that came with the dataset about citric acid adding ‘freshness and flavor’ to wine seems as though it may not apply to red wines. It certainly will be interesting to see how citric acid (or its absence) correlates with wine quality in this particular dataset. It’s also worth noting that most wines in this dataset do contain traces of citric acid even though these are all red wines.
Total Acid
Let’s compute the sum of the three forms of acid identified in the dataset to make a ‘total.acidity’ variable. I would hypothesize that this sum should have a very strong negative correlation with pH (see below) in which the lower the value, the stronger the acid.
Next, let’s look at the structure of the new variable.
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 5.270 7.827 8.720 9.118 10.070 17.045Glancing at the summary values, it looks like this is going to be right-skewed with a long tale at the higher values. Let’s plot a histogram to confirm.
Yep, that hypothesis was correct.
Residual Sugar
Residual sugar seems to have a very long tail, so let’s perform a log 10 transformation.
Still a long tail, but the middle of the distribution is more clearly normalish. Let’s try doing this a different way to see if we can eliminate the outliers.
That’s a little better.
Chlorides
This is the amount of salt (sodium chloride) found in the wine.
Chlorides definitely has a long-tail, so let’s try removing the outliers. I have never experienced salt in wine as a good thing, so it would also be interesting to examine the outliers on the high end of the scale.
This plot definitely is more normal in its appearance than the previous one.
Free Sulfur Dioxide
Total Sulfur Dioxide
Sulfur dioxide in both its free and bound form is used as a preservative (see above). Since the plots for both are right skewed, let’s perform a square root transformation on each.
First on the free sulfur dioxide:
And total sulfur dioxide
That mostly got rid of the right skew and the curves are more normal.
Density
Density is a quite a normal distribution with few outliers: the mean and median are almost identical.
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.9901 0.9956 0.9968 0.9967 0.9978 1.0037pH (Overall Acidity)
pH (overall acidity) also shows a normalish distribution with few outliers.
Sulphates
Sulphates, an additive which serves as a preservative, demonstrates a right skewed curve with a long tail. Let’s see what how it looks after a square root transformation.
That’s normalish.
Alcohol Content % By Volume
Let’s summarize the alcohol variable:
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 8.40 9.50 10.20 10.42 11.10 14.90Alcohol content by volume is slightly right-skewed and in the 9.5-11 range. This is fairly intutive to anyone who drinks wine. It will be interesting to look at correlations between high/low alcohol content and the quality score.
Univariate Analysis
What is the structure of your dataset?
The structure of the dataset is as discussed in the “Structure and Summary” section above. One major transformation we made in EDA was creating a factorized version of the quality variable. Another transformation was adding a variable - total.acidity - which is the sum of the three acid(ity) variables.
What is/are the main feature(s) of interest in your dataset?
The main feature of interest is the distribution of the quality variable. More than three-quarters of the observations are of the middle two ratings which means that only about 400 items were classified at either the upper or lower extremes. It should be possible to examine that group of ‘very good’ or ‘very bad’ wines and to correlate with other variables.
A few of the input variables appear to be related to each other: the variables related to acidity and to sulfur dioxide. It will be interesting to see how those items correlate with each other and with the quality score.
What other features in the dataset do you think will help support your investigation into your feature(s) of interest?
Biggest question for me is how/whether the different input variables correlate or predict the output variable, quality. Am particularly curious to see how alchohol level correlates with quality. Will also be interesting to see how the different acidity-related variables correlate with each other and with quality.
Did you create any new variables from existing variables in the dataset?
As mentioned previously, I did create a total.acidity variable by summing up three other acid(ity) variables and we also transformed two of the existing variables: (‘X’ and ‘quality’) into factor variables. We don’t need these variables to be summarized and it makes more intuitive sense for these variables to be factors.
Of the features you investigated, were there any unusual distributions?
Did you perform any operations on the data to tidy, adjust, or change the form
of the data? If so, why did you do this?
Yes, several of the variables’ plots displayed a skew or a relatively large number of outliers. I transformed these variables using base-10 logarithims, square root operations, or, in a couple cases, by cutting off the outliers. As a result, the transformed plots were easier to read and showed more normal distributions.
Bivariate Plots Section
To get started with the bivariate analysis, we will subset the dataframe to plot acid-related variables with each other and against quality. The rw_acid dataframe has the following structure:
## 'data.frame': 1599 obs. of 6 variables:
## $ quality : int 5 5 5 6 5 5 5 7 7 5 ...
## $ pH : num 3.51 3.2 3.26 3.16 3.51 3.51 3.3 3.39 3.36 3.35 ...
## $ total.acidity : num 8.1 8.68 8.6 12.04 8.1 ...
## $ fixed.acidity : num 7.4 7.8 7.8 11.2 7.4 7.4 7.9 7.3 7.8 7.5 ...
## $ volatile.acidity: num 0.7 0.88 0.76 0.28 0.7 0.66 0.6 0.65 0.58 0.5 ...
## $ citric.acid : num 0 0 0.04 0.56 0 0 0.06 0 0.02 0.36 ...Following is a plot of rw_acid (made using the ggpairs function of the ggally library). Note that we are using the integer quality variable so that the function can calculate correlation coefficients.
As expected, the correlations among the acid-related variables are quite high and in some cases approach 1.
The greatest correlations between acidity and quality are volatile acidity (negative correlation - medium) and citric acid (positive correlation - weak). Volatile acidity of wine is aceitic acid which at high quantities gives the taste of balsamic vinegar so the negative correlation is not surprising. Citric acid is a flavor enhancer in white wines, but as noted above is not allowed by the EU in red wine, so the positive correlation is a bit more surprising. Perhaps iron and copper are introduced via the grape crushing process and citric acid is used (as permitted) to remove their traces.
I was surprised at the nearly 100% correlation between fixed acidity and total acidity, but a visual inspection of the dataframe shows that I should not have been.
## quality pH total.acidity fixed.acidity volatile.acidity citric.acid
## 566 5 3.14 13.79 12.7 0.60 0.49
## 879 6 3.22 9.60 8.8 0.61 0.19
## 1205 7 3.40 8.02 7.2 0.36 0.46Fixed acidity is by far the largest component of total acidity in grams per volume (dm^3). Since the definition of a ‘fixed’ acid is that it does not easily evaporate, this makes sense. There is hardly any correlation between either total or fixed acidity and quality.
Alcohol and Residual Sugar
Let’s examine another conceptually related pair of variables: alchohol and residual.sugar. Since alcohol in wine is created during fermentation by chemical reaction with sugar in grapes, our hypothesis is the higher the level of alcohol, the lower the amount of residual sugar.
We again subset the dataframe in order to view the following scatterplot.
Let’s view a scatterplot of residual sugar vs. alchohol (faceted by quality). In order to make the high and low quality wines more visible, I filtered out the 75% of observations in the medium quality (5-6) range.
It actually does look as though the superior quality scores (8) tend to be higher in alcohol content (though a bit all over the place in terms of residual sugar). The wines that scored lower in quality (3-4) are grouped on the lower side of the alcohol/volume scale (and are also all over the place in terms of residual sugar).
Let’s look at the actual correlation for the entire sample between alcohol content and residual sugar.
## [1] 0.04207544Hypothesis here appears to be incorrect since the correlation coefficient is very small. A little Google research on wine yields some clues - see the following citations.
Since we’ve noted that fermentation converts grape sugars to alcohol, there’s definitely a relationship between a wine’s residual or unconverted sugar and its alcohol level. The more residual sugar, the lower the potential alcohol that wine could have. But it’s not a simple relationship—there are other factors at play, particularly how much sugar the grapes had to begin with. Two wines could have widely different alcohol percentages but the same amount of residual sugar—or vice versa—depending on how ripe the grapes were. Wine Spectator
So, while there’s definitely a loose relationship between a wine’s residual sugar and its alcohol level, it’s not a simple relationship. EverWonderWine? Blog
Since it is alcohol content that has the greater correlation with quality, we’ll use it in the subsequent plots.
Let’s examine another scatterplot matrix, this time using those variables which we suspect may be correlated with quality:
- volatile acidity
- citric acid
- chlorides
- sulphates
- alcohol
This time we’ll subset using the dplyr library.
The correlation coefficients are pretty informative; the plots somewhat less so. The input variables that correlate with quality are:
- volatile acidity - medium (negative) 0.391
- citric acid - low 0.226
- sulphates - low 0.251
- alcohol - medium 0.476
The input variables that correlate with other variables are:
- citric acid and volatile acidity - medium (negative) 0.552
- citric acid and chlorides - low 0.204
- citric acid and sulphates - medium 0.313
- chlorides and sulphates - medium 0.371
- sulphates and volatile acidity - low (negative) 0.261
- alcohol and chlorides - low (negative) 0.221
- alcohol and volatile acidity - low (negative) 0.202
Alright, I’m going to run the same scatterplot using the dataframe with the 5-6 quality range filtered out to see if anything stands out at the high and low ranges.
The correlations are slightly stronger for wines at higher and lower quality ratings which make the scatterplots easier to interpret. We will look at some of the multi- variate plots with the same filter.
Boxplots for Key Input Variables Across Quality
The following series of boxplots will visualize the input variables with higher correlation coefficients across the factorized Quality variable.
Alcohol and Quality
The correlation of higher alcohol/volume with higher quality is clearly seen here particularly at the higher quality ratings.
Alcohol/Volume Summarized by Quality Level
These are the summary stats which are depicted in the above boxplot.
## factor(rw_copy$quality): 3
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 8.400 9.725 9.925 9.955 10.575 11.000
## --------------------------------------------------------
## factor(rw_copy$quality): 4
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 9.00 9.60 10.00 10.27 11.00 13.10
## --------------------------------------------------------
## factor(rw_copy$quality): 5
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 8.5 9.4 9.7 9.9 10.2 14.9
## --------------------------------------------------------
## factor(rw_copy$quality): 6
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 8.40 9.80 10.50 10.63 11.30 14.00
## --------------------------------------------------------
## factor(rw_copy$quality): 7
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 9.20 10.80 11.50 11.47 12.10 14.00
## --------------------------------------------------------
## factor(rw_copy$quality): 8
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 9.80 11.32 12.15 12.09 12.88 14.00Only the wines rated “5” diverge from the general pattern.
Sulphates and Quality
The correlation between sulphates and quality is weak, but nevertheless visible.
Citric Acid and Quality
The correlation between citric acid and quality is also visible here. However the overlap in values indicated by the interquartile ranges and the substantial number of wines with no citric acid (discussed above) discount the correlation.
Volatile Acidity and Quality
The inverse correlation between volatile acidity and quality is visible here. This is the second-strongest correlation after alchohol.
Bivariate Analysis
Talk about some of the relationships you observed in this part of the investigation. How did the feature(s) of interest vary with other features in the dataset?
Much of this analysis is offered inline (see above above).
The two strongest relationships with quality variable are the positive correlation with alcohol and the negative one with volatile acidity. If I were to create a model predicting which items account for most of a high quality score it would be based on these two variables. Show me a vinho verde with high alcohol content per volume and low volatile acidity and I’ll bet you it’s highly-rated wine.
Did you observe any interesting relationships between the other features (not the main feature(s) of interest)?
Yes,the positive correlations among the various acids and acidity-related variables confirms the validity of the data. It was also interesting not to find an expected (negative correlation beteween alcohol and residual sugar level).
What was the strongest relationship you found?
The strongest relationship in this dataset is between alcohol and quality and also among the various acidity variables.
Multivariate Plots Section
Volatile Acidity and Alcohol Over Quality
These are the two strongest relationships with quality - volatile acidity is a negative correlation and alcohol is a positive one. Setting up alcohol on the x-axis and volatile acidity on the y-axis, I would expect to see most of the high quality wine in the lower right-hand portion of the plot and the lower-quality ones on the upper left. To make the visualization more comprhensible, I’m going to plot a subset excluding the 1200 values in the medium (5-6) quality range.
Bingo! This was exactly what we were expecting.
For completeness, here are same items plotted for the entire dataset.
This plot is harder to read, but if you examine it carefully, it shows the same trend.
Alcohol vs. Sulphates Over Quality
Sulphates (“a wine additive which acts as an antimicrobial and antioxidant”) correlates weakly with both quality and volatile acidity. Let’s plot the relationships among these three variables.
The relationship between alcohol/volume and quality is again clearly visible; the correlation between sulpahates, quality and alcohol is weak, but also visible.
Volatile Acidity vs. Sulphates Over Quality
Since sulphates and volatile acidity have a negative correlation and both have a relationship with quality, I though I might see a pretty clear negative slope in this plot. There is a slight negative slope indicating the correlation between the two input variables. And these seem to correlate with quality, but honestly, relationships are not as clear as I had hoped.
Multivariate Analysis
Talk about some of the relationships you observed in this part of the investigation. Were there features that strengthened each other in terms of looking at your feature(s) of interest?
We examined a number of the factors that ggpairs had shown to have relatively high correlation and plotted them with quality as a factor.
Were there any interesting or surprising interactions between features?
The plot of alcohol and volatile acidity against quality factors confirmed that these were the strongest correlations.
OPTIONAL: Did you create any models with your dataset? Discuss the strengths and limitations of your model.
I did not create any models with the dataset. This is something I would certainly do for the more polished analysis after the EDA phase. ——
Final Plots and Summary
Plot One
Histogram of the Quality Variable
And here is the summary distribution of that variable:
## 3 4 5 6 7 8
## 10 53 681 638 199 18Description for Plot One
Sometimes the most descriptive stats are the simplest count statistics. This is one of those cases: over 75% of the quality scores fall in the medium (5-6) range so the task of analyzing this data is going to learn what are the variables that drive the top and bottom ranges.
Plot Two
Boxplots of Alcohol and Volatile Acidity Faceted by Quality
Description Two
There are several ways to understand the correlation between quality and alchohol and quality and volatile acid, but in my view the box plots are the best visualization. Not only do they show that as the input variables change, the quality score increases or decreases, but the overlap in the ranges is also indicative.
Plot Three
Multivariate Scatterplot Alcohol/Volatile Acidity/Quality
## Scale for 'colour' is already present. Adding another scale for
## 'colour', which will replace the existing scale.
Description Three
This is the plot which shows most clearly the two greatest relationships with Quality as discussed above. Setting up alcohol on the x-axis and volatile acidity on the y-axis, I expected to see most of the high quality wine in the lower right-hand portion of the plot and the lower-quality ones on the upper left. The plot is easier to understand with the middle values removed.
Reflection
This project demonstrates the power of R and the GGPlot libarary to create visualizations that help an analyst to better understand a dataset before starting with advanced analytics. Using a pre-cleaned and tidied red wine dataset eliminated some of the difficulties inherent in data wrangling, but because the data is limited it also reduced the possibilities for interesting discovery.
While there are clear relationships among variables - notably alcohol, volatile acidity, and quality - this is the place to note the axiom “correlation does not imply causation.” Knowing what I know intuitively as a wine drinker myself, in particular I wouldn’t even generalize these relationships to all wines, but would focus further investigation on specific qualities of Vinho Verde. I don’t think it is true that higher alcohol wines are generally higher quality, but that seems to be the case within the Vinho Verde family and it would be worth exploring.
For further exploration, I would proceed with the following:
- I would build a linear model using the 3 or 4 variables that correlate with quality to understand the extent to which the combinations account for variation in quality.
- I would conduct a similar exploration using the white whine (vinho verde) dataset. It would be particularly interesting to see if the citric acid correlations are the stronger, the same or weaker since, in general, citric acid is much more prevalent and beneficial in white wines.
- I would look for a similar dataset from a different family of wines and, if such data could be obtained, I would run a similar EDA over it.